Early ictal extent did not predict seizure recurrence in 36 HUP SEEG patients

We tested whether a larger spatial extent of early ictal involvement predicts seizure recurrence after surgery or ablation in HUP SEEG patients. The preregistered analysis used archived per-seizure, per-bipolar-pair features for 37 patients and 104 seizures from OpenNeuro ds004100 v1.1.3, with clinician-marked onset, and archived outcome labels. The primary label-free measure was the fraction of bipolar pairs with a 30–100 Hz power rise above 3.2-fold in the first 10 seconds. Patient values were means across seizures. Among 36 patients with known outcomes (16 not seizure-free, 20 seizure-free; 101 seizures), AUC P(extent in not seizure-free > extent in seizure-free) was 0.375 (95% CI 0.19–0.57; permutation p = 0.21; Holm-adjusted p = 0.42). None of four tests was significant; all AUC estimates were below 0.5, but the opposite direction was not significant. The upper primary confidence bound was below 0.72, so these data do not support an effect of AUC 0.72 or greater in the predicted direction. Sensitivity analyses did not change the conclusion. Extent was stable within patients (ICC(1) = 0.87). With 36 patients, effects near AUC 0.6 remain possible. The measure is a spectral proxy on implanted pairs, not the Epileptogenicity Index; onset artifacts, limited sampling, at most three seizures per patient, predominantly laser ablation, and unknown follow-up time limit interpretation.

epilepsy · stereo-EEG · ictal onset · seizure outcome · signal analysis · surgical outcome · OpenNeuro ds004100